HP HSTNS-PD43 PSU Hack & Feedback Circuit (Partial)

[Trurl]

I wanted to hack a high output server PSU for use during automotive ECU programming, and found this HSTNS-PD43 PSU to be a good candidate especially after finding a simple hack to boost output voltage and adjust the OVP setting.


For a better understanding of the PSU's feedback circuit I traced out the auxiliary PCB shown in the video (just the fully accessible side of the double sided/multi-layered PCB, as the backside is mostly blocked by the large main/high voltage cap) and put it in schematic form below.

HSTNS-PD43 Auxiliary PCB (Fully Accessible Side)
Photo


Schematic (Large*)
*Composite image of 8 monitor screenshots: I suggest open(click) the schematic thumbnail below, then right-click on opened schematic > "Save Image As" and reopen downloaded schematic in your preferred image viewing app, for easier viewing(zooming, jumping around etc.).


*Notes*
1. Schematic is in grid form (alphabetical rows / numeric columns). The primary side columns are marked with added "P" and are numbered right to left (to differentiate from secondary side which was completed first).

2. Main output voltage feedback points(vias) are marked "+12V+" and their simulation voltages reflect the set power rail voltage which can be other than 12V, hence the added volt meter(which is obviously not present on the PCB) to show the active set power rail voltage (e.g. 15.5V, 14V, 13.8V etc.).

3. Vias and hidden traces of which connections were confirmed, if not indicated with a visible trace, have been marked with "terminals"(line ending in a circle) and marked with the terminal's grid code, followed by the destination's device pin(s) or another grid code in parentheses - "( )" [e.g. C10(PIC-21/22) suggests, terminal in C10 connects with PIC pins 21/22].

4. Vias or trace ends who's destinations have not been determined are marked with a "?" followed by their grid locations (e.g. ?E13)

5. All caps are uniformly marked 1uF as they cannot be easily determined without desoldering and doing so would be a pain as they are so tiny. Also, the cap symbol on the schematic is for a polarized cap, but obviously those on this auxiliary PCB are NOT(please ignore the suggestion of polarization - my bad).

6. The first 1/3rd on the left is isolated from the 2/3rds on the right.

7. Ground(GND) points are marked with measured resistance values* with respect to the main output ground tab or in the primary section's case, the negative pin of the main high voltage cap(cap voltage @ 0.0V).
*e.g. 0R, 0R1, 0R2 for 0.0, 0.1, 0.2Ohm etc. - a few are higher and may not be "direct" links to ground.

[The Left Side]
1. TI UCD3138 (Digital Controller)
2. TI LMR12010 (Switching Regulator marked "SF7B")

[The Right Side]
1. Microchip dsPIC33FJ64GS606
2. TI UCC3895 (Phase-Shift PWM Controller)
3. TI LMR12010(Switching Regulator marked "SF7B")
4. Several OpAmps/Comparators

[Backside Of PCB]
There are more ICs and 3 more opto-couplers.
         

The 3 connectors in the 1st of 3 photos above from left to right are for AC input control(black), fan, and status LED. They were disconnected for better views of the backside.

[Opto-Coupler On Mainboard]
I even spotted one on the main PCB  "hiding out" between the big cap and the auxiliary PCB.


[Daughter Board]
Has numerous OnSemi FDMS039N08B N-Channel POWERTRENCH MOSFETs.


[Simulation Notes Regarding Voltage and OVP Hack]
Note: Only "passive" components(resistors, capacitors, and single diodes) were active during simulation(all "active" semiconductors were "excluded" from simulation as I do not have access to the software driving the controls, and therefore opamps/comparators would not function as they are intended).

1. In "normal" 12V operation, the 12V feedback voltage is reduced to about 2.7V at the initial comparator's negative input(LM324A pin-2, in D8), and the OVP related voltage to about 4V at LM393 pin-3(in D12).


2. When the hack resistors(12k & 4.3k as in the video) are connected, the 2.7V drops to about 1.8V, and the 4V drops to about 1.6V. Note: this is with the simulation power rail kept at 12.3V.


3. If the simulation power rail is adjusted to 15.5V output(as in the unit in the video), the 2.7V drops to 2.27V, and the 4V drops to 2.07V.


Higher valued resistors drop the FB voltages less so lower voltage boosting is possible.


If there are any issues with the schematic, I'd appreciate comments, and will update the schematic above in this first post so it is always the updated version.

Cheers~

[Trurl]

I've moved the questions from the first post to this 2nd post (for better visibility), and added some more below.

[Questions]
Despite having observed all of the above, I'm still baffled about the following:

1. What type of feedback compensator is used in this PSU's feedback circuit and where are its core traces/components on the schematic? Perhaps a "type 3" mutant not in textbooks?

2. How is the OVP being controlled and where are its core traces/components on the schematic?

3. What are the roles of the 2 switching regulators marked "SF7B" (Texas Instruments LMR12010) - 1 on the primary side and 1 on the secondary side?
* In low cost PSUs, a single switching regulator like the "SF7B" is used as the "brains" for control, and its FB(Feed Back) input's voltage divider resistors can be manipulated for voltage adjustment. But the PD43 PSU has relatively advanced ICs controlling this switching regulator(and 2 of them!), so I'm wondering how these 2 regulators are "specialized" for the digital controller on the left primary side, and for the PIC on the right secondary side.

Thanks...

[Trurl]

I've tested the HSTNS-PD43 PSU's OVP triggering with a range of OVP mod resistances over a range of voltages and relatively low power loads (6 ~ 2 Ohms). The objective was to get an idea of how this PSU's OVP triggers to determine the optimal resistance for keeping the OVP function working (instead of completely disabling it), and for it to trigger just below 16V which is its output caps' rating.

I've organized the test results into the chart (spreadsheet image) below:

HSTNS-PD43 Modified OVP Triggering Results


OVP mod resistances between 10.50k ~ 10.60k resulted in the OVP mostly triggering at 15.50V and just below(15.40V) but did not trigger below. Hence, this range would be the best resistances for keeping OVP active while being able to use the widest range of output voltages.
Please note though that I've not yet been able to test with high loads (e.g. 50~100A), to determine OVP triggering voltages at such loads. I will test this in the future once I get setup with a sufficient nichrome wire dummy load and update.

Above 15.50V, the voltage regulation is jumpy (i.e. doesn't like to be set to 15.60V and will either jump to say 15.63V or 15.53V etc.). The OVP triggering test results were also more unpredictable than at lower voltages.

* Notes About The Chart's Markings:
1. Cell Colors
- Pink: Mixed results (i.e. both triggering and not triggering at a set voltage, depending on load). OVP not triggering where desired or triggering where it shouldn't. Best to avoid.
- Yellow: Mixed results that could be acceptable, but may require further testing and/or sufficient consideration.
- Green: Mostly uniform & consistent results(i.e. all triggering or all not triggering). OVP triggering and not triggering as preferred (i.e, triggering at 15.50V and above while not trigger below).

2. The loads were 6, 5, 4, 3, 2 Ohm power resistors resulting in roughly 2~7A loading with slight differences between the different voltages set.

3. "X" = No Trigger, "O" = Trigger (A number to the right of these marks indicates consecutively repeated results (i.e. "X3" = No Trigger 3 times consecutively).

4. Each "X" or "O" marking represents 1 second held loading x4 times or fast tapping up to x20 times(or until trigger), and the hold vs fast tapping are distinguished with a slash(i.e. "/") in between (i.e. "X/O" = no trigger with hold x4 / trigger with fast tapping up to x20).

5. Red markings indicate unexpected unusual results.

6. Blue markings indicate the few test results from briefly using a wire wound pot for the OVP mod resistor. All results were "X"(OVP never triggers!) likely due to the wire wound pot's inductance messing with the OVP feedback voltage.

7. "TAPO" = (OVP) Triggers At Power ON: 12.66V x 15.00k and 15.50V x 20.35k combinations resulted in the PSU's OVP triggering at power ON.

8. Single marked cells were the very first round of test results in which hold & tapping were done randomly as opposed to by strict counting, and the resistance was infrequently checked(for temperature influenced drift). Hence, resistance drift may have lowered the accuracy of these results.

9. Multi marked cells, especially with markings in brackets (i.e. "[ ]") were recorded with much care - frequently checking the resistances (acceptable drift was kept within +/- 0.01k), the latest results in black brackets being checked before and after testing for EVERY marking.

10. For any other details, please refer to the top of the attached chart above.

Cheers...

[Nik094]

I have some questions about this hack. Im trying to run 2 in series as a lifepo4 24v charger. They both power on in series and Iv done the hack on both as well as eliminating ground to board on one of them. They both power on in series no problem and I had them pushing 78amps at 29.48v enough to charge battery to 27v. But one kept whining and hissing and turning off. So I pulled the loud one apart and turns out my ovp resistor on one the solder joint didn’t hold. So I fixed that! Go out plug them back in no more noise. But when I plug them into load (302ah 29.2v battery) and they turn off. Any ideas what it could be? I used the resistor from your comment on the YouTube video and used 8k resistors for ovp hack.

[Trurl]

I have not personally tried linking 2 units as you are doing so I can only point to those that likely have. In the YouTube video's comments, one of the earliest comments (from "1 year ago", at the time of this writing),
'Mr06Postman' commented,
"Thanks for this nate. As some one else asked, do you know which pin is for load sharing yet?"
And the OP - "Hopper"(hopperbuiltamps2135) replied,
"Bottom of power supply. 5th little pin from the end."
Keeping this in mind, your most likely method of getting some useful tips might be to reply to Mr06Postman's comment on the YouTube video's 'Comments' section.

Regarding your 'load', I wonder if the 29.2V 302AH rating is just too much for the PD43 PSU to handle as it is rated as 1400W, and I assume at 29.2V the max output of a dual setup would be   just under 96A? (1400Wx2 = 2800W, 2800W/29.2V = 95.89A) This of course is assuming that any load sharing setup(referred to above) is also correctly setup. This might be an OCP(Over Current Protection) kicking in, and not the OVP, in your case. With no current limiting circuit between the PSUs and the battery, I think the  battery could be pulling way more than the PSUs can supply.

Regarding OVP resistor, I've not yet been able to test OVP triggering at very high loads (near 100A) and that's why I made a note in my previous post regarding this. You might want to try a smaller value resistance such as the 4.3k that Hopper suggested for his 15.5V setup.

But I have a feeling that the load is just too large and the OCP is kicking in to prevent the PSU from a meltdown.

Look forward to hearing about your testing results! Cheers...

P.S.
One other thing came to mind, you might try checking the voltage of the battery before hooking up the PSUs, and making sure that the PSUs are set to a slightly higher voltage before connecting them to the battery and turning ON the PSUs, to prevent reverse flow of current. It might be necessary to connect a sufficiently high amp rated switch(in OFF state) on the negative side, before connecting the PSUs to the battery, and turn ON the PSUs first for like a minute or so to fully charge up all caps etc., with the switch(on the negative side) in OFF state, then after the initial minute or so of warmup, turn the switch ON. This would ensure forward current from the PSUs to the battery.

[Nik094]

The switch part at the end was my thought as it seems like a back current issue almost. I did have it charging at 79amps but it was hissing as one of the ovp mods solder did not take. It stopped entirely once I fixed it lol. I’m going to add a remote 600a relay switch on negative side today that I had sitting around from a golf cart build and try smaller resistors.

[Trurl]

I have a 100A rated "kill switch"(like those used on enthusiast racing "track cars") on hand for my future build. This allows me to have the cables connected to the car chassis' power terminals which eventually connect to the battery in the trunk, but have the negative to the PSU disconnected(switched OFF) say during PSU warmup, and during this time I can switch my modular volt meter to show the car's battery voltage status then the PSU's voltage setting and adjust as needed(to have the PSU's voltage slightly higher than the car's battery) before turning ON the "kill switch" which will be connected to the PSU's negative(ground) side.

While testing the OVP triggering (at the relatively low loads of 6~2 Ohms, resulting in about 2~7A loading) I noticed that triggering sometimes was different between tests done with say 20 seconds of warmup and 60 seconds of warmup. I could hear the fan speed changing (increasing) as well from around  20~30 seconds, and again around 45~60 seconds.

As I was at some points triggering very frequently in the higher voltage settings, I was able to notice that the PSU's OVP tends to be more trigger sensitive when it has had little warm up time. For example, with only 20seconds of warm up, the OVP could trigger at the very first tap of switching where as with a longer say 50~60 second warm up time, the OVP might trigger only after say the 3rd or 4th tapped switching ON, or in some cases not trigger when the PSU locks into "high current mode"(which is my own observation, not an official specification feature that I'm aware of) In this "mode" the fan speed his relatively high/very steady and no loading (whether steady holds or irregular fast tapping) will knock it off balance. This behavior I figure is associated with the PSU having charged up(or not) its main high voltage cap, as well as any other small caps that were drained in the previous OVP triggering event.

Although I will only need to use one of these PSUs in up to around 14.6 V / 50~90A range, I look forward to your experience with your setup and would appreciate your update here or in the YouTube video's comments section, particularly how if at all you incorporate the "load sharing" feature(using the PSU's pin 5, underside from edge) for your setup. Cheers~

[zheka64]

Hello. I'm curious about how to implement a current limit for these power supplies. Given that chargers based on this power supply are sold in Ukraine, it's clearly feasible. Attached are some examples of devices available on our local marketplace.

[Trurl]

If such units with "current limiting" modifications are not too over-priced, perhaps you could purchase one, open the enclosure to see how the seller achieved the "currrent limitiing", and share your findings here. I've yet to encounter any info on the matter online.

Other than the OVP mod resistances of interest that I've posted above, the voltage mod resistances in my case are as follows:

[ Resistor Values for 12.66V ~ 15.50V ]
* As tested without rounding. Used a mix of metal film resistors and a wire wound potentiometer on a breadboard.

12.66V  -  100.1k~103.5k
12.75V  -  80.1k
12.80V  -  71.0k
12.89V  -  59.83k
13.00V  -  50.63k
13.10V  -  44.20k
13.20V  -  39.09k
13.30V  -  35.09k
13.40V  -  31.82k
13.50V  -  29.23k
13.60V  -  26.92k
13.70V  -  24.99k
13.80V  -  23.36k
13.90V  -  21.95k
14.00V  -  20.64k
14.10V  -  19.46k
14.20V  -  18.43k
14.30V  -  17.48k
14.40V  -  16.62k
14.50V  -  15.87k
14.60V  -  15.20k
14.70V  -  14.54k
14.80V  -  14.01k
14.90V  -  13.46k
15.00V  -  12.91k~12.98k
15.10V  -  12.45k
15.20V  -  12.05k
15.30V  -  11.64k
15.40V  -  11.27k
15.50V  -  10.68k

Cheers~

[zheka64]

The price is quite overpriced compared to the price of the power supply unit itself. Different options cost $100-140. I would probably buy it if I hadn't already purchased a charger based on the HSTNS-PL11.